CN113268403B - Time series analysis and prediction method, device, equipment and storage medium - Google Patents

Time series analysis and prediction method, device, equipment and storage medium Download PDF

Info

Publication number
CN113268403B
CN113268403B CN202110572876.3A CN202110572876A CN113268403B CN 113268403 B CN113268403 B CN 113268403B CN 202110572876 A CN202110572876 A CN 202110572876A CN 113268403 B CN113268403 B CN 113268403B
Authority
CN
China
Prior art keywords
time sequence
time
time series
data
value
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202110572876.3A
Other languages
Chinese (zh)
Other versions
CN113268403A (en
Inventor
金叶
徐锐
沈松
王柯
朱威
张皞
武晓頔
李东佩
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China United Network Communications Group Co Ltd
Original Assignee
China United Network Communications Group Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China United Network Communications Group Co Ltd filed Critical China United Network Communications Group Co Ltd
Priority to CN202110572876.3A priority Critical patent/CN113268403B/en
Publication of CN113268403A publication Critical patent/CN113268403A/en
Application granted granted Critical
Publication of CN113268403B publication Critical patent/CN113268403B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00—Error detection; Error correction; Monitoring
    • G06F11/30—Monitoring
    • G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
    • G06F11/3447—Performance evaluation by modeling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

The application provides a time sequence analysis and prediction method, a device, equipment and a storage medium, wherein the method provides a simple time sequence analysis and prediction method, a preset analysis and prediction model is used for predicting a critical value of time sequence data of a target server, the preset analysis and prediction model is obtained by training the time sequence data of a reference server and the critical value of the time sequence data of the reference server at preset time, and further, the prediction critical value of the time sequence data is obtained, wherein the time sequence does not need to reach a steady state in the prediction process, the problem that the existing model construction process is complicated and the time sequence is required to be a steady non-white noise time sequence when the time sequence analysis and prediction is performed by using a differential autoregressive moving average model is solved.

Description

Time series analysis and prediction method, device, equipment and storage medium
Technical Field
The present application relates to the field of equipment operation and maintenance technologies, and in particular, to a method, an apparatus, a device, and a storage medium for time series analysis and prediction.
Background
And (3) continuously developing random economy, and setting more and more server resources in the process of floor cloud service by each large operator. Therefore, how to efficiently maintain these servers and efficiently utilize all server resources is a major concern and difficulty for large operators.
In the related art, during long-term operation of the server, various load monitoring indexes, such as the utilization rate of a central processing unit (Central Processing Unit, CPU), the utilization rate of a memory, etc., generate a large amount of time-series data. The technician uses an analysis prediction method to perform analysis prediction on the time series data, for example, analysis prediction on a critical value (such as a CPU utilization maximum value, a memory utilization minimum value, etc.) of the next moment. Therefore, technicians can guide operation and maintenance work of the server based on the analysis and prediction results, and the server resources can be utilized as efficiently as possible. The analysis and prediction method mainly adopted by the prior art is a time sequence analysis and prediction method based on a differential autoregressive moving average model.
However, the construction process of the differential autoregressive moving average model is complicated, and the differential autoregressive moving average model requires that the time series be a stationary non-white noise time series. If the time series cannot reach a plateau, a differential autoregressive moving average model cannot be used. Therefore, how to provide a simple time series analysis and prediction method, and to make subsequent analysis and prediction based on the time series regardless of whether the time series reaches a steady state, becomes an urgent problem to be solved.
Disclosure of Invention
In order to solve the problems in the prior art, the application provides a time sequence analysis and prediction method, a device, equipment and a storage medium.
In a first aspect, an embodiment of the present application provides a method for analyzing and predicting a time sequence, including the steps of:
acquiring time sequence data of a target server, wherein the time sequence data comprises one or more of CPU (Central processing Unit) utilization time sequence, memory utilization time sequence, average input/output (IO) request time sequence, network card receiving byte number per second time sequence and network card sending byte number per second time sequence;
inputting the time series data of the target server into a preset analysis prediction model, wherein the preset analysis prediction model is obtained by training the time series data of a reference server and a critical value corresponding to the time series data of the reference server at preset time, and the critical value comprises a maximum value and/or a minimum value;
and obtaining a prediction critical value of the time series data of the target server according to the output of the preset analysis prediction model.
In one possible implementation manner, before the time series data of the target server is input into a preset analysis prediction model, the method further includes:
Preprocessing the time series data of the target server, wherein the preprocessing comprises One or more of data deduplication, time series set division, data type conversion, missing value filling, time series feature extraction, time series feature construction, one-Hot encoding (One-Hot encoding) and data fusion;
the inputting the time series data of the target server into a preset analysis prediction model comprises the following steps:
and inputting the preprocessed time series data into the preset analysis and prediction model.
In one possible implementation, the preprocessing includes the time series feature extraction and the time series feature construction;
the preprocessing the time series data of the target server comprises the following steps:
extracting a first sequence feature of the time sequence data of the target server and a second sequence feature of historical time sequence data corresponding to the time sequence data of the target server;
determining values of preset feature variables according to the first sequence features and the second sequence features, wherein the preset feature variables comprise one or more of a day of the week, a day of the work, a weekend and a holiday;
And carrying out time sequence feature construction according to the value of the preset feature variable.
In one possible implementation manner, after the obtaining, according to the output of the preset analysis prediction model, a prediction threshold value of the time series data of the target server, the method further includes:
acquiring an actual critical value corresponding to the time sequence data of the target server;
and determining the prediction accuracy of the preset analysis prediction model according to the actual critical value and the prediction critical value.
In a possible implementation manner, the determining the prediction accuracy of the preset analysis prediction model according to the actual critical value and the prediction critical value includes:
calculating a difference value between the actual critical value and the predicted critical value;
and determining the prediction accuracy of the preset analysis prediction model according to the difference value.
In one possible implementation manner, after the determining the prediction accuracy of the preset analysis prediction model according to the actual critical value and the prediction critical value, the method further includes:
if the prediction accuracy is lower than a preset accuracy threshold, generating an unreliable prompt of the preset analysis prediction model, and retraining the preset analysis prediction model according to the time sequence data of the reference server and a critical value of the time sequence data of the reference server, which corresponds to the critical value of the time sequence data in the preset time.
In a second aspect, an embodiment of the present application provides a time-series analysis and prediction apparatus, including:
the acquisition module is used for acquiring time sequence data of the target server, wherein the time sequence data comprises one or more of a CPU (Central processing Unit) utilization rate time sequence, a memory utilization rate time sequence, an average IO (input/output) request time sequence, a network card receiving byte number per second time sequence and a network card transmitting byte number per second time sequence;
the input module is used for inputting the time series data of the target server into a preset analysis prediction model, wherein the preset analysis prediction model is obtained through training of the time series data of a reference server and a critical value corresponding to the time series data of the reference server at preset time, and the critical value comprises a maximum value and/or a minimum value;
and the prediction module is used for obtaining the prediction critical value of the time series data of the target server according to the output of the preset analysis prediction model.
In one possible implementation manner, the input module is specifically configured to:
preprocessing the time sequence data of the target server, wherein the preprocessing comprises One or more of data deduplication, time sequence set division, data type conversion, missing value filling, time sequence feature extraction, time sequence feature construction, one-Hot encoding and data fusion;
And inputting the preprocessed time series data into the preset analysis and prediction model.
In one possible implementation, the preprocessing includes the time series feature extraction and the time series feature construction;
the input module is specifically configured to:
extracting a first sequence feature of the time sequence data of the target server and a second sequence feature of historical time sequence data corresponding to the time sequence data of the target server;
determining values of preset feature variables according to the first sequence features and the second sequence features, wherein the preset feature variables comprise one or more of a day of the week, a day of the work, a weekend and a holiday;
and carrying out time sequence feature construction according to the value of the preset feature variable.
In a possible implementation manner, the system further includes an evaluation module, configured to obtain an actual critical value corresponding to the time series data of the target server after the prediction module obtains a predicted critical value of the time series data of the target server according to the output of the preset analysis prediction model;
and determining the prediction accuracy of the preset analysis prediction model according to the actual critical value and the prediction critical value.
In one possible implementation, the evaluation module is specifically configured to:
calculating a difference value between the actual critical value and the predicted critical value;
and determining the prediction accuracy of the preset analysis prediction model according to the difference value.
In one possible implementation, the evaluation module is further configured to:
if the prediction accuracy is lower than a preset accuracy threshold, generating an unreliable prompt of the preset analysis prediction model, and retraining the preset analysis prediction model according to the time sequence data of the reference server and a critical value of the time sequence data of the reference server, which corresponds to the critical value of the time sequence data in the preset time.
In a third aspect, an embodiment of the present application provides a time-series analysis-prediction apparatus, including:
a processor;
a memory; and
a computer program;
wherein the computer program is stored in the memory and configured to be executed by the processor, the computer program comprising instructions for performing the method according to the first aspect.
In a fourth aspect, embodiments of the present application provide a computer-readable storage medium storing a computer program, the computer program causing a server to execute the method of the first aspect.
In a fifth aspect, embodiments of the present application provide a computer program product comprising computer instructions for performing the method of the first aspect by a processor.
According to the method, the device, the equipment and the storage medium, time series data of a target server are obtained, the time series data comprise one or more of CPU utilization rate time series, memory utilization rate time series, average IO request time series, network card receiving byte number per second time series and network card sending byte number per second time series, further, the time series data of the target server are input into a preset analysis prediction model, according to the output of the preset analysis prediction model, a prediction critical value of the time series data of the target server is obtained, the preset analysis prediction model is obtained through training of the time series data of a reference server and the critical value of the reference server corresponding to the time series data at preset time, the critical value comprises a maximum value and/or a minimum value, namely the embodiment of the application provides a simple time series analysis prediction method, the critical value of the time series data of the target server is predicted through the preset analysis prediction model, and further, the prediction critical value of the time series data is obtained, the prediction critical value of the time series data in a steady state regression time series is achieved through a steady state, and the problem that the existing time series is not required to be stably analyzed is solved, and the time series is not required to be even in a steady state.
Drawings
In order to more clearly illustrate the embodiments of the application or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the description below are only some embodiments of the application, and that other drawings can be obtained according to these drawings without inventive faculty for a person skilled in the art.
FIG. 1 is a schematic diagram of a time-series analysis and prediction system architecture according to an embodiment of the present application;
FIG. 2 is a schematic flow chart of a time series analysis and prediction method according to an embodiment of the present application;
FIG. 3 is a schematic diagram of missing value filling according to an embodiment of the present application;
FIG. 4 is a schematic diagram of performing time-series analysis prediction by using a preset analysis prediction model according to an embodiment of the present application;
FIG. 5 is a flowchart of another method for analyzing and predicting a time sequence according to an embodiment of the present application;
fig. 6 is a schematic structural diagram of a time-series analysis and prediction apparatus according to an embodiment of the present application;
FIG. 7 is a schematic structural diagram of another time-series analysis and prediction apparatus according to an embodiment of the present application;
FIG. 8A provides one possible basic hardware architecture of the time-series analysis prediction device of the present application;
fig. 8B provides another possible basic hardware architecture of the time-series analysis-prediction device of the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
The terms "first," "second," "third," and "fourth" and the like in the description and in the claims and in the above drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the application described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
In the related art, during the long-term operation of the server, various load monitoring indexes, such as CPU utilization, memory utilization, etc., may generate a large amount of time-series data. The technician uses an analysis prediction method to perform analysis prediction on the time series data, for example, analysis prediction on a critical value (such as a CPU utilization maximum value, a memory utilization minimum value, etc.) of the next moment. Therefore, technicians can guide operation and maintenance work of the server based on the analysis and prediction results, and the server resources can be utilized as efficiently as possible.
The analysis and prediction method mainly adopted by the prior art is a time sequence analysis and prediction method based on a differential autoregressive moving average model. Based on a differential autoregressive moving average model: the model requires that the time sequence is a stable non-white noise time sequence, and the stability of the time sequence is checked by an ADF unit root stability check method or a drawing observation method, and then the white noise check is performed on the time sequence by an Ljung-Box autocorrelation check method. For an unstable time sequence, the model needs to obtain a stable time sequence through d-level differential operation or logarithmic operation. And for a stable non-white noise sequence, the model obtains p, d and q values of the model through an autocorrelation function and a partial autocorrelation function, so that subsequent time sequence analysis prediction is carried out.
However, the construction process of the above differential autoregressive moving average model is complicated, and the differential autoregressive moving average model requires that the time series be a stationary non-white noise time series. If the time series cannot reach a plateau, a differential autoregressive moving average model cannot be used.
In order to solve the above problems, an embodiment of the present application provides a simple time series analysis and prediction method, which predicts a critical value of time series data of a target server by presetting an analysis and prediction model, and further obtains a prediction critical value of the time series data, wherein the time series does not need to reach a steady state in the prediction process, and the problems that the existing time series analysis and prediction using a differential autoregressive moving average model is complex in model construction process and requires that the time series is a steady non-white noise time series are solved.
Alternatively, the method for analyzing and predicting a time sequence provided by the present application may be applied to a schematic architecture of a system for analyzing and predicting a time sequence shown in fig. 1, where the system may include at least one of a receiving device 101, a processing device 102 and a display device 103 as shown in fig. 1.
In a specific implementation process, the receiving device 101 may be an input/output interface or a communication interface, and may be configured to receive time-series data of the target server, where the time-series data includes one or more of a CPU usage time series, a memory usage time series, an average IO request number time series, a network card receiving byte number per second time series, and a network card transmitting byte number per second time series.
The processing device 102 may obtain the time series data of the target server through the receiving device 101, or may directly obtain the time series data of the target server from a database of an operator, and further predict a critical value of the time series data of the target server through the preset analysis prediction model, to obtain a prediction critical value of the time series data, where the prediction process does not need to reach a steady state, so as to solve the problem that the existing time series analysis prediction using a differential autoregressive moving average model has complicated model construction process and requires that the time series be a steady non-white noise time series.
The display device 103 may be configured to display time-series data of the target server, a prediction threshold value of the time-series data, and the like.
The display device may also be a touch display screen for receiving user instructions while displaying the above content to enable interaction with a user.
It should be understood that the above-described processor may be implemented by a processor that reads instructions in a memory and executes the instructions, or may be implemented by a chip circuit.
The above system is only one exemplary system, and may be set according to application requirements when implemented.
It should be understood that the structure illustrated in the embodiments of the present application does not constitute a specific limitation on the architecture of the analysis and prediction system for time series. In other possible embodiments of the present application, the architecture may include more or less components than those illustrated, or some components may be combined, some components may be split, or different component arrangements may be specifically determined according to the actual application scenario, and the present application is not limited herein. The components shown in fig. 1 may be implemented in hardware, software, or a combination of software and hardware.
In addition, the system architecture and the service scenario described in the embodiments of the present application are for more clearly describing the technical solution of the embodiments of the present application, and do not constitute a limitation on the technical solution provided by the embodiments of the present application, and as a person of ordinary skill in the art can know, along with the evolution of the system architecture and the appearance of the new service scenario, the technical solution provided by the embodiments of the present application is also applicable to similar technical problems.
The following description of the present application is given by taking several embodiments as examples, and the same or similar concepts or processes may not be described in detail in some embodiments.
Fig. 2 is a schematic flow chart of a time series analysis and prediction method, which may be implemented by any device that performs the time series analysis and prediction method, and the device may be implemented by software and/or hardware. As shown in fig. 2, based on the system architecture shown in fig. 1, the analysis and prediction of the time sequence provided by the embodiment of the present application may include the following steps:
s201: and acquiring time sequence data of the target server, wherein the time sequence data comprises one or more of a CPU (Central processing Unit) utilization rate time sequence, a memory utilization rate time sequence, an average IO request time sequence, a network card receiving byte number per second time sequence and a network card transmitting byte number per second time sequence.
In the embodiment of the present application, taking the execution subject as an example of the processing device in fig. 1, the processing device acquires time series data of a target server, where the target server may be a server that needs to perform analysis and prediction on a time series of the server, and may be determined according to an actual situation.
For example, the processing device may obtain, from the Hive database of the operator, time-series data of the target server, for example, time-series data of 5 indexes (CPU usage, memory usage, average IO request number, number of bytes received by the network card per second, and number of bytes transmitted by the network card per second) for 2 months, where the time-series data may include the following fields: acquisition date (data_date), acquisition hour (data_hour), server IP (IP), maximum value (max_value), minimum value (min_value), average value (avg_value), and the like. For example, as shown in table 1, a time series of indicators (e.g., CPU usage, memory usage, average IO requests, number of bytes received per second by the network card, or number of bytes sent per second by the network card) includes 5 features: acquisition date, acquisition hour, maximum, minimum and average.
TABLE 1
Date of collection Collecting hours Maximum value Minimum value Average value of
2019/8/26 0 42.9 0.59 18.695
S202: inputting the time series data of the target server into a preset analysis prediction model, wherein the preset analysis prediction model is obtained by training the time series data of a reference server and a critical value corresponding to the time series data of the reference server at preset time, and the critical value comprises a maximum value and/or a minimum value.
The processing device may input the time-series data of the target server into a preset analysis prediction model, and obtain a prediction threshold value of the time-series data of the target server according to an output of the preset analysis prediction model.
Here, the time-series data of the reference server may be the time-series data of any known server, and the threshold value at the preset time corresponding to the time-series data of the reference server may be an actual threshold value at the preset time corresponding to the time-series data of the reference server. The preset time may be determined according to a time situation, for example, the acquisition date of the time series data of the reference server may be 2019/8/26, the preset time may be 2019/8/27, that is, a critical value at the preset time corresponding to the time series data of the reference server may be a critical value at 2019/8/27 corresponding to the time series data of the reference server.
In addition, in order to improve accuracy of analysis prediction of a subsequent time series, the processing device may perform preprocessing on the time series data of the target server, where the preprocessing includes One or more of data deduplication, time series set partitioning, data type conversion, missing value filling, time series feature extraction, time series feature construction, one-Hot encoding, and data fusion, before inputting the time series data of the target server into a preset analysis prediction model.
Illustratively, the preprocessing described above includes data deduplication. The processing device may detect whether or not the time-series data of the target server has duplicate data, and if so, perform deduplication processing on the time-series data of the target server, before inputting the time-series data of the target server into a preset analysis prediction model.
When detecting whether the time-series data of the target server has duplicate data, the processing device may detect the time-series data based on a field included in the time-series data, for example, a collection date and a collection hour. If the same acquisition date and acquisition hour exist in the time series data of a certain index (CPU utilization rate, memory utilization rate, average IO request times, network card received byte number per second and network card transmitted byte number per second), the processing device judges that repeated data exist and performs deduplication processing.
For example, as shown in table 2, if the acquisition date and the acquisition hour are the same in two time series of one index, the processing device determines that there is duplicate data.
TABLE 2
Date of collection Collecting hours Maximum value Minimum value Average value of
2019/8/26 0 42.9 0.59 18.695
2019/8/26 0 42.9 0.1 17.555
In an embodiment of the present application, the preprocessing may further include time series set partitioning. The processing device may divide the time series set according to time series data of different indexes (CPU usage, memory usage, average IO request number, number of bytes received by the network card per second and number of bytes transmitted by the network card per second) before inputting the time series data of the target server into a preset analysis prediction model, for example, the time series data of one person index is divided into one time series set. The processing device may acquire the IP of the target server, and divide time-series data of the plurality of servers based on the IP of the server and the different index. For example, let the IP of the server be i1, i2, i3, … …, and the index ID of the CPU usage, the memory usage, the average IO request number, the number of bytes received by the network card per second, and the number of bytes transmitted by the network card per second be x1, x2, x3, x4, x5. When ip=i1 and id=x1, time-series data of a certain index (for example, CPU utilization) of a server is obtained and divided into a time-series set.
Here, the preprocessing may further include data type conversion. The type of the data required to be input into the model by the preset analysis prediction model is a preset type, and the processing device may detect whether the type of the time-series data of the target server is a preset type before inputting the time-series data of the target server into the preset analysis prediction model, and if not, perform data type conversion. From the above, the time series data of one index includes the fields: acquisition date, acquisition hour, maximum, minimum, average, etc. The types of the different fields may be different, and the preset types may include various types. For example, the preset type corresponding to the acquisition date may be a date type, and the preset type corresponding to the maximum value may be float64 or the like. The processing device detects whether the acquisition date in the time-series data of the target server is date type, whether the maximum value is float64, and the like, and if not, the data type is converted.
The preprocessing described above may also include missing value padding. In order to continue the data in the subsequent processing, the processing device may fill the time-series data of the target server with a missing value before inputting the time-series data of the target server into a preset analysis prediction model, for example, when the value on a certain day is entirely missing, the processing device may fill the time-series data with a mode filling method; when the value on a day is partially missing, it is filled with 0. As shown in fig. 3, the processing device detects whether the data on the T-th day is missing entirely, calculates the date of T-30 if the data is missing entirely, and determines whether the data in the (T-30, T) section exists. If present, the data for the H hours on day T is filled with the mode for all H hours in the (T-30, T) interval. If not, the data is filled with 0 on day T, hours H. In addition, if the data on the T day is not missing entirely, the processing device detects whether the data on the H hour on the T day is missing, and if so, the data on the H hour is filled with 0.
The preprocessing may further include time series feature extraction and time series feature construction. The processing device may extract a first sequence feature of the time-series data of the target server and a second sequence feature of historical time-series data corresponding to the time-series data of the target server before inputting the time-series data of the target server into a preset analysis prediction model, and further determine a value of a preset feature variable including one or more of a day of the week, a weekend, and a holiday according to the first sequence feature and the second sequence feature, thereby performing time-series feature construction according to the value of the preset feature variable. The historical time series data may be determined according to practical situations, for example, the historical time series data of n hours before the last m days of the day.
For example, the processing device may make the collection dates d1, d2, …, di, …, d60, and the collection hours h0, h1, …, hj, … (0 < = h0, h1, … hj, …, h23< = 23), where h0=0, h1=1, h2=2, … …, and the method for extracting features from a certain index is as follows:
For each piece of data t=di, h=hj, extracting data Q hours (1 < =p < =30, 0< =q < =24) before the day P days before the day expands to the characteristics of the day at the hour, and the data set at the hj on the day di after expansion consists of the following two parts:
(1) Original features (first sequence features) at hj on day di: { di, hj, max_value, min_value, avg_value };
(2) Features n hours before the first m days on day hj (second sequence features) (i-P < = m < = i-1, j-Q < = n < = j, m, n take all integers throughout the interval): { di-m, hj-n, max_value (i-m, j-n), min_value (i-m, j-n), avg_value (i-m, j-n) }.
On the basis of the extracted features, the processing device continues to perform feature construction, and the feature construction method is as follows:
let the variable day identify that the day is a day of the week, day e { x|1< = x < = 7}, the variable work_day identify that the day is a workday, work_day e {0,1}, the variable week_day identify that the day is a weekend, weeky_day e {0,1}, the variable holiday identify that the day is a holiday, holiday e {0,1}, and the day sequence d1, d2, …, di, …, d60 is used to calculate when t=di that the T day is a day of the week, is a workday, is a weekend, is a holiday. The data construction method on day t=di is as follows:
Body features were constructed on day t=di: day of week, day of weekday, day of weekend, day of holiday, i.e., { daydi, work_daydi, week_daydi, holidaydi }
Body features were constructed on day t=di-1: yesterday is the day of week, yesterday is the weekday, yesterday is the weekend, yesterday is the holiday, i.e., { day di-1, work_day di-1, weekday_day di-1, holidaydi-1}
Body features on day t=di+1 were constructed: whether tomorrow is a day of the week, whether tomorrow is a weekday, whether tomorrow is a holiday, i.e., { daydi+1, work_daydi+1, weekDaydi+1, holidaydi+1}
Features of di, di-1, di+1 days were combined to form a dataset on day t=di:
{daydi-1,work_daydi-1,week_daydi-1,holidaydi-1,daydi,work_daydi，week_daydi,holidaydi,daydi+1,work_daydi+1,week_daydi+1,holidaydi+1}
in addition, the preprocessing can also comprise One-Hot encoding and data fusion. Before the time series data of the target server is input into a preset analysis prediction model, the processing device can further expand the data set by using One-Hot coding, wherein the method for expanding the processing device is as follows:
the data fields which need to be subjected to One-Hot coding in the data set are selected, and the data fields mainly comprise the following fields: { T, H, dm, hn, daydi-1, work_daydi-1, week_daydi-1, holidaydi-1, daydi, work_daydi, week_daydi, holidaydi, daydi+1, work_daydi+1, week_daydi+1, holidaydi+1}, wherein 1< = i < = 60,0< = j < = 23, i-P < = m < = i-1, j-Q < = n < = j,1< = P < = 30,0< = Q < = 24.
Then, the fields are classified, and different One-Hot coding modes are adopted for different categories. The dataset fields are mainly divided into 4 classes: date, hour, week, boolean. The One-Hot encoding modes of these categories are as follows:
for the date class field, a field can be expanded into a 31-dimensional value, the number of the day is the number of the day with the dimension marked 1, and the rest of the dimensions take 0. The number of the date fields is P+1, and the date fields are expanded into (P+1) multiplied by 31 dimensions after being encoded by One-Hot. For example, when p=4, the date class field is extended from 5 dimensions to 155 dimensions.
For the hours class field, one field can be extended to a 24-dimensional value, and the current hour is taken as 1 in the dimension, and the rest hours are taken as 0. The total number of the hour fields is P multiplied by Q+1, and the number of the hour fields is expanded into (P multiplied by Q+1) multiplied by 24 after One-Hot coding. For example, when p=4, q=4, the hours class field is extended from 17 dimensions to 408 dimensions.
For the week field, one field can be extended to a 7-dimensional value, and the day is the day of the week, and the other hours take 1 in the dimension of the day and 0 in the other hours. The number of the peripheral fields is P+1, and the peripheral fields are expanded into (P+1) x 7 dimensions after being encoded by One-Hot. For example, when p=4, the week field is extended from 5 dimensions to 35 dimensions.
For a boolean field, one field may be extended to a 2-dimensional value. The number of Boolean fields is 9, and the Boolean fields are expanded into 9X 2 dimensions after One-Hot encoding.
After the expansion, the method for the processing device to perform data fusion is as follows:
the original data set is divided into 5 indexes according to the prop_id field, and is divided into 5 data blocks. For each metric data block, it is further divided into 24 hour data blocks according to the data_hour field. For each hour data block, the data block is divided into single data according to the date, and one piece of data corresponds to the data of a certain hour of a certain day. I.e. the data of the data block of the h hour is the data set of all h hours in the time series. The One-Hot encoding is performed on the prop_id field.
S203: and obtaining the prediction critical value of the time series data of the target server according to the output of the preset analysis prediction model.
For example, as shown in fig. 4, the processing device may acquire the time series data of the target server, input the time series data of the target server, and further obtain the prediction threshold of the time series data of the target server by using the preset analysis prediction model.
After obtaining the prediction threshold value of the time series data of the target server, the processing device can guide the operation and maintenance work of the server according to the prediction threshold value, and realize the efficient utilization of the server resources as much as possible.
According to the embodiment of the application, the time series data of the target server are obtained by obtaining one or more of a CPU utilization rate time series, a memory utilization rate time series, an average IO request time series, a network card receiving byte number per second time series and a network card transmitting byte number per second time series, and then the time series data of the target server are input into a preset analysis prediction model, and according to the output of the preset analysis prediction model, the prediction critical value of the time series data of the target server is obtained, wherein the preset analysis prediction model is obtained by training the critical value at preset time corresponding to the time series data of the reference server, and the critical value comprises a maximum value and/or a minimum value.
In addition, in the embodiment of the present application, after obtaining the prediction threshold value of the time-series data of the target server according to the output of the preset analysis prediction model, the processing device further considers evaluating the preset analysis prediction model. Fig. 5 is a flowchart of another time series analysis and prediction method according to an embodiment of the present application. As shown in fig. 5, the method includes:
s501: and acquiring time sequence data of the target server, wherein the time sequence data comprises one or more of a CPU (Central processing Unit) utilization rate time sequence, a memory utilization rate time sequence, an average IO request time sequence, a network card receiving byte number per second time sequence and a network card transmitting byte number per second time sequence.
S502: inputting the time series data of the target server into a preset analysis prediction model, wherein the preset analysis prediction model is obtained by training the time series data of a reference server and a critical value corresponding to the time series data of the reference server at preset time, and the critical value comprises a maximum value and/or a minimum value.
S503: and obtaining the prediction critical value of the time series data of the target server according to the output of the preset analysis prediction model.
Steps S501 to S503 are described in the above steps S201 to S203, and are not described herein.
S504: and acquiring an actual critical value corresponding to the time series data of the target server.
S505: and determining the prediction accuracy of the preset analysis prediction model according to the actual critical value and the prediction critical value.
Here, the processing device may calculate a difference between the actual threshold value and the predicted threshold value, and determine a prediction accuracy of the preset analysis prediction model according to the difference.
For example, the processing device may calculate a Root Mean Square Error (RMSE) using the following formula after calculating the difference between the actual threshold value and the predicted threshold value:
where n is the number of test samples, yi is the actual threshold value obtained,is the predictive threshold of the model.
And the processing device determines the prediction accuracy of the preset analysis prediction model according to the RMSE. The smaller the RMSE, the smaller the difference between the predicted value and the actual value, the higher the model accuracy, and the more accurate the model prediction.
In addition, after determining the prediction accuracy of the preset analysis prediction model according to the actual critical value and the prediction critical value, the processing device may generate an untrusted prompt of the preset analysis prediction model if the prediction accuracy is lower than a preset accuracy threshold, and retrain the preset analysis prediction model according to the time series data of the reference server and the critical value of the preset time corresponding to the time series data of the reference server.
The preset accuracy threshold may be determined according to an actual situation. When the prediction accuracy of the preset analysis and prediction model is low, the processing device generates a corresponding prompt, and trains the preset analysis and prediction model again, so that the subsequent analysis and prediction of the time sequence are carried out by using the retrained model, and the accuracy of subsequent processing is improved.
In the embodiment of the present application, after obtaining the prediction threshold value of the time series data of the target server according to the output of the preset analysis prediction model, the processing device further considers evaluating the preset analysis prediction model to perform corresponding optimization based on the evaluation result, for example, if the accuracy of the evaluation model is low, training the model again, so as to improve the accuracy of the subsequent processing. In addition, the embodiment of the application provides a simple time sequence analysis and prediction method, wherein the preset analysis and prediction model is used for predicting the critical value of the time sequence data of the target server, so as to obtain the prediction critical value of the time sequence data, wherein the time sequence does not need to reach a stable state in the prediction process, and the problems that the existing time sequence analysis and prediction is carried out by using a differential autoregressive moving average model, the existing model construction process is complicated, and the time sequence is required to be a stable non-white noise time sequence are solved.
Fig. 6 is a schematic structural diagram of a time series analysis and prediction apparatus according to an embodiment of the present application, corresponding to the time series analysis and prediction method of the above embodiment. For convenience of explanation, only portions relevant to the embodiments of the present application are shown. Fig. 6 is a schematic structural diagram of a time-series analysis and prediction apparatus according to an embodiment of the present application, where the time-series analysis and prediction apparatus 60 includes: an acquisition module 601, an input module 602, and a prediction module 603. The time-series analysis/prediction device may be the processing device itself, or a chip or an integrated circuit for realizing the functions of the processing device. It should be noted that, the division of the acquisition module, the input module and the prediction module is only a division of a logic function, and both may be integrated or independent physically.
The acquiring module 601 is configured to acquire time-series data of the target server, where the time-series data includes one or more of a CPU utilization time series, a memory utilization time series, an average IO request number time series, a network card receiving byte number per second time series, and a network card sending byte number per second time series.
The input module 602 is configured to input the time series data of the target server into a preset analysis prediction model, where the preset analysis prediction model is obtained by training the time series data of a reference server and a critical value corresponding to the time series data of the reference server at a preset time, where the critical value includes a maximum value and/or a minimum value.
And a prediction module 603, configured to obtain a prediction threshold of the time series data of the target server according to an output of the preset analysis prediction model.
In one possible implementation, the input module 602 is specifically configured to:
preprocessing the time sequence data of the target server, wherein the preprocessing comprises One or more of data deduplication, time sequence set division, data type conversion, missing value filling, time sequence feature extraction, time sequence feature construction, one-Hot encoding and data fusion;
and inputting the preprocessed time series data into the preset analysis and prediction model.
In one possible implementation, the preprocessing includes the time series feature extraction and the time series feature construction;
The input module 602 is specifically configured to:
extracting a first sequence feature of the time sequence data of the target server and a second sequence feature of historical time sequence data corresponding to the time sequence data of the target server;
determining values of preset feature variables according to the first sequence features and the second sequence features, wherein the preset feature variables comprise one or more of a day of the week, a day of the work, a weekend and a holiday;
and carrying out time sequence feature construction according to the value of the preset feature variable.
The device provided by the embodiment of the present application may be used to implement the technical scheme of the embodiment of the method described in fig. 2, and its implementation principle and technical effects are similar, and the embodiment of the present application is not described here again.
Fig. 7 is a schematic structural diagram of another time-series analysis and prediction apparatus according to an embodiment of the present application, and based on fig. 6, the time-series analysis and prediction apparatus 60 further includes: an evaluation module 604.
In a possible implementation manner, the evaluation module 604 is configured to obtain, after the prediction module obtains the prediction threshold value of the time series data of the target server according to the output of the preset analysis prediction model, an actual threshold value corresponding to the time series data of the target server;
And determining the prediction accuracy of the preset analysis prediction model according to the actual critical value and the prediction critical value.
In one possible implementation, the evaluation module 604 is specifically configured to:
calculating a difference value between the actual critical value and the predicted critical value;
and determining the prediction accuracy of the preset analysis prediction model according to the difference value.
In one possible implementation, the evaluation module 604 is further configured to:
if the prediction accuracy is lower than a preset accuracy threshold, generating an unreliable prompt of the preset analysis prediction model, and retraining the preset analysis prediction model according to the time sequence data of the reference server and a critical value of the time sequence data of the reference server, which corresponds to the critical value of the time sequence data in the preset time.
The device provided by the embodiment of the present application may be used to implement the technical scheme of the embodiment of the method described in fig. 5, and its implementation principle and technical effects are similar, and the embodiment of the present application is not described here again.
Alternatively, fig. 8A and 8B schematically provide one possible basic hardware architecture of the time-series analysis-prediction device of the present application, respectively.
Referring to fig. 8A and 8B, the time series analysis prediction apparatus includes at least one processor 801 and a communication interface 803. Further optionally, a memory 802 and a bus 804 may also be included.
Wherein the number of processors 801 may be one or more in the time series analysis prediction apparatus, fig. 8A and 8B illustrate only one of the processors 801. Optionally, the processor 801 may be a CPU, a graphics processor (Graphics Processing Unit, GPU) or digital signal processing (Digital Signal Process, DSP). If the time-series analysis prediction apparatus has a plurality of processors 801, the types of the plurality of processors 801 may be different or may be the same. Alternatively, the plurality of processors 801 of the time-series analysis-and-prediction apparatus may be integrated as a multi-core processor.
Memory 802 stores computer instructions and data; the memory 802 may store computer instructions and data required to implement the above-described time series analysis and prediction method provided by the present application, for example, the memory 802 stores instructions for implementing the steps of the above-described time series analysis and prediction method. Memory 802 may be any one or any combination of the following storage media: nonvolatile Memory (e.g., read-Only Memory (ROM), solid State Drive (SSD) or Solid State Drive, hard Disk Drive (HDD), optical Disk), and volatile Memory.
The communication interface 803 may provide an information input/output to the at least one processor. Any one or any combination of the following devices may also be included: a network interface (e.g., ethernet interface), a wireless network card, etc., having network access functionality.
Optionally, the communication interface 803 may also be used for data communication with other computing devices or terminals by a time-series analysis prediction device.
Further alternatively, fig. 8A and 8B represent bus 804 with a bold line. Bus 804 may connect processor 801 with memory 802 and communication interface 803. Thus, through bus 804, processor 801 may access memory 802 and may also interact with other computing devices or terminals using communication interface 803.
In the present application, the time-series analysis-prediction apparatus executes the computer instructions in the memory 802, so that the time-series analysis-prediction apparatus implements the above-described time-series analysis-prediction method provided by the present application, or so that the time-series analysis-prediction apparatus deploys the above-described time-series analysis-prediction device.
From a logical functional partitioning perspective, as illustrated in fig. 8A, the memory 802 may include an acquisition module 601, an input module 602, and a prediction module 603. The inclusion herein is not limited to physical structures, but rather involves only the functionality of the acquisition module, the input module, and the prediction module, respectively, as the instructions stored in memory are executed.
For example, as shown in FIG. 8B, an evaluation module 604 may also be included in the memory 802. The inclusion herein is not limited to a physical structure, but rather involves only the execution of instructions stored in memory to perform the functions of the evaluation module.
In addition, the time-series analysis and prediction apparatus described above may be implemented in hardware as a hardware module or as a circuit unit, in addition to the software as in fig. 8A and 8B described above.
The present application provides a computer-readable storage medium storing a computer program that causes a server to execute the above-described time-series analysis prediction method provided by the present application.
The present application provides a computer program product comprising computer instructions for execution by a processor of the method of analysis and prediction of a time series provided by the present application.
The application provides a chip comprising at least one processor and a communication interface providing information input and/or output for the at least one processor. Further, the chip may also include at least one memory for storing computer instructions. The at least one processor is configured to invoke and execute the computer instructions to perform the above-described method of predicting time series analysis provided by the present application.
In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in hardware plus software functional units.

Claims (9)

1. A method for analysis and prediction of a time series, comprising:
acquiring time sequence data of a target server, wherein the time sequence data comprises one or more of a central processing unit utilization rate time sequence, a memory utilization rate time sequence, an average input/output request time sequence, a network card receiving byte number per second time sequence and a network card transmitting byte number per second time sequence;
inputting the time series data of the target server into a preset analysis prediction model, wherein the preset analysis prediction model is obtained by training the time series data of a reference server and a critical value corresponding to the time series data of the reference server at preset time, and the critical value comprises a maximum value and/or a minimum value;
obtaining a prediction critical value of the time sequence data of the target server according to the output of the preset analysis prediction model;
before the time series data of the target server is input into a preset analysis prediction model, the method further comprises:
preprocessing the time series data of the target server, wherein the preprocessing comprises time series feature extraction, time series feature construction and single-heat coding;
The preprocessing the time series data of the target server comprises the following steps:
extracting time sequence characteristics of the time sequence data of the target server;
carrying out time sequence feature construction on the time sequence data after the time sequence feature extraction;
carrying out field selection on the time sequence data after the time sequence feature construction by utilizing the single-hot coding, and classifying the selected fields; the fields are classified into 4 classes: date, hour, week, boolean;
wherein, the single-hot coding modes of the fields of different categories are as follows:
for the date class field, a field is expanded to a 31-dimensional value; the number of the date fields is P+1, and the date fields are expanded into (P+1) multiplied by 31 after single-heat encoding;
for the hours class field, one field is extended to a 24-dimensional value; the number of the hour fields is P multiplied by Q+1, and the number of the hour fields is expanded into (P multiplied by Q+1) multiplied by 24 after single thermal coding;
for the week field, one field is extended to a 7-dimensional value; the number of the peripheral fields is P+1, and the peripheral fields are expanded into (P+1) x 7 dimensions after single thermal coding;
for the boolean field, one field is extended to a 2-dimensional value; the number of the Boolean fields is 9, and the number of the Boolean fields is expanded into 9 multiplied by 2 dimensions after single thermal coding; wherein 1< = P < = 30,0< = Q < = 24;
The preprocessing further comprises type conversion; the time series data includes fields: acquisition date, acquisition hour, maximum value, minimum value, average value;
detecting whether the acquisition date in the time series data of the target server is a date type, and whether the maximum value is float64; if not, data type conversion is performed.
2. The method of claim 1, wherein the step of determining the position of the substrate comprises,
the inputting the time series data of the target server into a preset analysis prediction model comprises the following steps:
and inputting the preprocessed time series data into the preset analysis and prediction model.
3. The method of claim 2, wherein the preprocessing the time series data of the target server comprises:
extracting a first sequence feature of the time sequence data of the target server and a second sequence feature of historical time sequence data corresponding to the time sequence data of the target server;
determining values of preset feature variables according to the first sequence features and the second sequence features, wherein the preset feature variables comprise one or more of a day of the week, a day of the work, a weekend and a holiday;
And carrying out time sequence feature construction according to the value of the preset feature variable.
4. A method according to any one of claims 1 to 3, further comprising, after the obtaining of the prediction threshold value of the time-series data of the target server according to the output of the preset analytical prediction model:
acquiring an actual critical value corresponding to the time sequence data of the target server;
and determining the prediction accuracy of the preset analysis prediction model according to the actual critical value and the prediction critical value.
5. The method of claim 4, wherein determining the prediction accuracy of the pre-determined analytical prediction model based on the actual threshold value and the predicted threshold value comprises:
calculating a difference value between the actual critical value and the predicted critical value;
and determining the prediction accuracy of the preset analysis prediction model according to the difference value.
6. The method of claim 4, further comprising, after said determining a prediction accuracy of said pre-determined analytical prediction model based on said actual threshold value and said predicted threshold value:
if the prediction accuracy is lower than a preset accuracy threshold, generating an unreliable prompt of the preset analysis prediction model, and retraining the preset analysis prediction model according to the time sequence data of the reference server and a critical value of the time sequence data of the reference server, which corresponds to the critical value of the time sequence data in the preset time.
7. An analysis and prediction apparatus for time series, comprising:
the system comprises an acquisition module, a storage module and a processing module, wherein the acquisition module is used for acquiring time sequence data of a target server, and the time sequence data comprises one or more of a central processing unit utilization rate time sequence, a memory utilization rate time sequence, an average input/output request time sequence, a network card receiving byte number per second time sequence and a network card transmitting byte number per second time sequence;
the input module is used for inputting the time series data of the target server into a preset analysis prediction model, wherein the preset analysis prediction model is obtained through training of the time series data of a reference server and a critical value corresponding to the time series data of the reference server at preset time, and the critical value comprises a maximum value and/or a minimum value;
the prediction module is used for obtaining a prediction critical value of the time sequence data of the target server according to the output of the preset analysis prediction model;
the input module is specifically configured to: preprocessing the time series data of the target server, wherein the preprocessing comprises time series feature extraction, time series feature construction and single-heat coding;
The preprocessing the time series data of the target server comprises the following steps:
extracting time sequence characteristics of the time sequence data of the target server;
carrying out time sequence feature construction on the time sequence data after the time sequence feature extraction;
carrying out field selection on the time sequence data after the time sequence feature construction by utilizing the single-hot coding, and classifying the selected fields; the fields are classified into 4 classes: date, hour, week, boolean;
wherein, the single-hot coding modes of the fields of different categories are as follows:
for the date class field, a field is expanded to a 31-dimensional value; the number of the date fields is P+1, and the date fields are expanded into (P+1) multiplied by 31 after single-heat encoding;
for the hours class field, one field is extended to a 24-dimensional value; the number of the hour fields is P multiplied by Q+1, and the number of the hour fields is expanded into (P multiplied by Q+1) multiplied by 24 after single thermal coding;
for the week field, one field is extended to a 7-dimensional value; the number of the peripheral fields is P+1, and the peripheral fields are expanded into (P+1) x 7 dimensions after single thermal coding;
for the boolean field, one field is extended to a 2-dimensional value; the number of the Boolean fields is 9, and the number of the Boolean fields is expanded into 9 multiplied by 2 dimensions after single thermal coding; wherein 1< = P < = 30,0< = Q < = 24;
The preprocessing further comprises type conversion; the time series data includes fields: acquisition date, acquisition hour, maximum value, minimum value, average value;
detecting whether the acquisition date in the time series data of the target server is a date type, and whether the maximum value is float64; if not, data type conversion is performed.
8. An analytic predictive device of a time series, comprising:
a processor;
a memory; and
a computer program;
wherein the computer program is stored in the memory and configured to be executed by the processor, the computer program comprising instructions for performing the method of any of claims 1-6.
9. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program, which causes a server to perform the method of any one of claims 1-6.
CN202110572876.3A 2021-05-25 2021-05-25 Time series analysis and prediction method, device, equipment and storage medium Active CN113268403B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110572876.3A CN113268403B (en) 2021-05-25 2021-05-25 Time series analysis and prediction method, device, equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110572876.3A CN113268403B (en) 2021-05-25 2021-05-25 Time series analysis and prediction method, device, equipment and storage medium

Publications (2)

Publication Number Publication Date
CN113268403A CN113268403A (en) 2021-08-17
CN113268403B true CN113268403B (en) 2023-10-31

Family

ID=77232800

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110572876.3A Active CN113268403B (en) 2021-05-25 2021-05-25 Time series analysis and prediction method, device, equipment and storage medium

Country Status (1)

Country Link
CN (1) CN113268403B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113515751A (en) * 2021-09-13 2021-10-19 富通云腾科技有限公司 Deployment analysis platform based on modular software
CN116415476A (en) * 2021-12-30 2023-07-11 维沃移动通信有限公司 Model construction method and device and communication equipment
CN115546493B (en) * 2022-12-06 2023-05-02 成都智元汇信息技术股份有限公司 Method, device and medium for filtering and weighing package identification information based on column data

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106886485A (en) * 2017-02-28 2017-06-23 深圳市华傲数据技术有限公司 Power system capacity analyzing and predicting method and device
CN109614231A (en) * 2018-12-04 2019-04-12 广东亿迅科技有限公司 Idle server resource discovery method, device, computer equipment and storage medium
CN109685290A (en) * 2019-02-11 2019-04-26 南方电网科学研究院有限责任公司 A kind of electricity demand forecasting method, device and equipment based on deep learning
CN110059858A (en) * 2019-03-15 2019-07-26 深圳壹账通智能科技有限公司 Server resource prediction technique, device, computer equipment and storage medium
CN111221706A (en) * 2020-01-06 2020-06-02 北京思特奇信息技术股份有限公司 CPU utilization rate prediction method, system, medium and equipment
CN111352733A (en) * 2020-02-26 2020-06-30 北京奇艺世纪科技有限公司 Capacity expansion and reduction state prediction method and device
CN111638958A (en) * 2020-06-02 2020-09-08 中国联合网络通信集团有限公司 Cloud host load processing method and device, control equipment and storage medium

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106886485A (en) * 2017-02-28 2017-06-23 深圳市华傲数据技术有限公司 Power system capacity analyzing and predicting method and device
CN109614231A (en) * 2018-12-04 2019-04-12 广东亿迅科技有限公司 Idle server resource discovery method, device, computer equipment and storage medium
CN109685290A (en) * 2019-02-11 2019-04-26 南方电网科学研究院有限责任公司 A kind of electricity demand forecasting method, device and equipment based on deep learning
CN110059858A (en) * 2019-03-15 2019-07-26 深圳壹账通智能科技有限公司 Server resource prediction technique, device, computer equipment and storage medium
CN111221706A (en) * 2020-01-06 2020-06-02 北京思特奇信息技术股份有限公司 CPU utilization rate prediction method, system, medium and equipment
CN111352733A (en) * 2020-02-26 2020-06-30 北京奇艺世纪科技有限公司 Capacity expansion and reduction state prediction method and device
CN111638958A (en) * 2020-06-02 2020-09-08 中国联合网络通信集团有限公司 Cloud host load processing method and device, control equipment and storage medium

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于LSTM的CPU资源使用情况预测算法的研究;樊文杰 等;《电子设计工程》;第29卷(第8期);23-28 *

Also Published As

Publication number Publication date
CN113268403A (en) 2021-08-17

Similar Documents

Publication Publication Date Title
CN113268403B (en) Time series analysis and prediction method, device, equipment and storage medium
CN110554958B (en) Graph database testing method, system, device and storage medium
CN110806954B (en) Method, device, equipment and storage medium for evaluating cloud host resources
CN108255897B (en) Visualized chart data conversion processing method and device
CN104036105B (en) It determines and is related to the method and system of the correctness of randomness application of big data analysis
CN108492005B (en) Project data processing method and device, computer equipment and storage medium
CN110147470B (en) Cross-machine-room data comparison system and method
CN104993962A (en) Method and system for obtaining use state of terminal
US20230066703A1 (en) Method for estimating structural vibration in real time
CN112906375A (en) Text data labeling method, device, equipment and storage medium
CN115730605B (en) Data analysis method based on multidimensional information
EP3336778A1 (en) Policy introduced effect prediction apparatus, policy introduced effect prediction program, and policy introduced effect prediction method
JP6658507B2 (en) Load estimation system, information processing device, load estimation method, and computer program
CN110232130B (en) Metadata management pedigree generation method, apparatus, computer device and storage medium
JP2023036681A (en) Task processing method, processing device, electronic equipment, storage medium, and computer program
CN116361567B (en) Data processing method and system applied to cloud office
CN107644042B (en) Software program click rate pre-estimation sorting method and server
CN109597702B (en) Root cause analysis method, device, equipment and storage medium for message bus abnormity
CN115344495A (en) Data analysis method and device for batch task test, computer equipment and medium
CN111124791A (en) System testing method and device
JP2023029604A (en) Apparatus and method for processing patent information, and program
CN113656292B (en) Multi-dimensional cross-space-time basic software performance bottleneck detection method
CN111666201A (en) Regression testing method, device, medium and electronic equipment
CN113656391A (en) Data detection method and device, storage medium and electronic equipment
CN113110804A (en) Duplicate picture deleting method, device, equipment and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant